Impact of age, follicle stimulating hormone and Johnsen's score on successful sperm retrieval by microdissection testicular sperm extraction

Abstract

Background:  The spermatozoa retrieval rate achieved by microdissection testicular sperm extraction (TESE) for patients with non‐obstructive azoospermia is still approximately 50%, despite technical advances. Predicting a successful outcome is important to avoid unnecessary surgery. A study was conducted to determine a method for calculating the predicted probability of successful microdissection TESE.

Methods:  Testicular spermatozoa were retrieved in 41 of 100 patients by microdissection TESE. Ten clinical factors were examined in the search for an optimal logistic model for predicting a good outcome. Once the best delimiting value was established, sensitivity and specificity were calculated.

Results:  Patient age, serum follicle stimulating hormone (FSH) concentration and Johnsen's score (JS) were shown to be influential factors by multivariate logistic analysis. The mathematical model for predicting the probability of a successful outcome was as follows: P = (1 + exp[0.144 × patient age − 0.059 × serum FSH concentration − 1.310 × JS])⁻¹. When a predictive probability of 49.7% was selected as the best cut‐off value, sensitivity and specificity were 78.0% and 76.3%, respectively.

Conclusion:  This present model is very useful for predicting successful microdissection TESE. In addition, we suggest that a younger patient age would increase the probability of success with this procedure. (Reprod Med Biol 2005; 4: 53–57)

INTRODUCTION

MATURE TESTICULAR SPERMATOZOA can be found in cases of non‐obstructive azoospermia (NOA). ¹ , ² A new microsurgical testicular sperm extraction (microdissection TESE) procedure that uses direct visualization with an operating microscope can target the large whitish tubules that presumably contain the greatest number of spermatogenetically active germ cells. ³ , ⁴ Many researchers have reported improved spermatozoa retrieval rates using this microsurgical technique. However, published data, including ours, have shown that spermatozoa retrieval rates achieved by microdissection TESE are still approximately 50%, despite this technical improvement. ⁴ , ⁵ , ⁶ , ⁷ , ⁸ This indicates that prediction of successful microdissection TESE is very important to avoid unnecessary surgery for half the patients with NOA. Recently, we reported a useful method to calculate the predicted probability of a successful outcome for microdissection TESE; P = (1 + exp[5.201 – 0.048 × serum follicle stimulating hormone (FSH) concentration − 0.449 × serum total testosterone (TT) concentration − 0.021 × serum inhibin B concentration])⁻¹. ⁹ This equation was obtained from multivariate logistic regression analysis of several ‘non‐invasive’ variables.

It is well known that testicular histologic findings are quite useful in predicting successful TESE, if they can be evaluated preoperatively. Despite this advantage, a diagnostic testicular biopsy is not generally recommended, because a preoperative testicular biopsy might cause inflammatory changes, hematoma, parenchymal fibrosis or permanent devascularizaion of the testes. ¹⁰ We do not perform a diagnostic testicular biopsy, either. However, as some patients are referred to our clinics for microdissection TESE after prior failed conventional TESE, they already have testicular samples, which can be evaluated histologically, without diagnostic biopsy before microdissection TESE. In addition, we obtain a small testicular sample for pathological examination at the time of microdissection TESE in all patients. Taken together, we suggest that the predictive model would be improved and very useful, especially for patients who consider microdissection TESE after prior failed conventional TESE in the future, if histological findings at the time of microdissection TESE are used for predictive analysis, instead of them obtained by preoperative diagnostic biopsy. Therefore, we conducted a retrospective study based on univariate and multivariate regression analyzes of preoperative ‘non‐invasive’ variables with histologic findings, to develop a new model for calculating the predicted probability of a successful for microdissection TESE in patients with NOA.

SUBJECTS AND METHODS

ONE HUNDRED PATIENTS who underwent microdissection TESE at Osaka University Hospital, Osaka Central Hospital, or Osaka Police Hospital were included in this retrospective study. All of these 100 patients were included in our previous study. ⁹ Patients were diagnosed with NOA on the basis of a complete history, physical examination, and endocrinologic profile before undergoing TESE and sperm freezing. Excluded from this study were patients with chromosomal abnormalities. Mean ± standard deviation age of patients was 33.6 ± 4.9 years, and the duration of infertility was 51.5 ± 44.5 months. Testicular spermatozoa were retrieved successfully in 41 of the 100 patients (41%). The microdissection TESE procedure was as described elsewhere. ⁸ In brief, small samples (10–15 mg) were excised from the larger, more opaque tubules under a surgical microscope at × 20 to × 40 magnification. The procedure was terminated when spermatozoa were retrieved or further biopsy was thought likely to jeopardize the blood supply of the testis.

Patient age, testicular volume as determined by orchidometer, mean Johnsen's score (JS) ¹¹ as determined by histological examination after TESE, and certain endocrinologic data were obtained. The endcrinologic data included concentrations of luteinizing hormone (LH), FSH, TT, free testosterone (FT) as measured by analog ligand radioimmunoassay, prolactin (PRL), estoradiol (E2), and inhibin B. These 10 variables were included in univariate logistic regression analysis. A b ₁ value as regression coefficient less than 0.10 was considered significant in the univariate model. We calculated correlation coefficients among the 10 factors and found that LH and FT were highly correlated with FSH and TT, respectively. Thus, excluding LH and FT, we used the remaining eight factors as explanatory variables to construct a multivariate logistic model for predicting the probability of a successful outcome. We then tried to construct an optimum logistic model with a stepwise procedure by setting the significance level at 0.20 and entering explanatory variables into the model that would maintain significances at this level. The odds ratio and 95% confidence interval were calculated as an estimate of the size of an effect. The probability of a successful outcome was estimated for each patient by means of the optimum model. To investigate the usefulness of the model, the best cut‐off value for the predicted probability to discriminate success from failure for microdissection TESE was determined on the basis of receiver operating characteristic (ROC) analysis. Once the best delimiting value was established, optimum sensitivity, specificity, and the area under the ROC curve (AUC^ROC) were calculated. All statistical analyzes were performed with the SAS Statistical Package (version 8.02; SAS Institute, Cary, NC, USA).

RESULTS

THE RESULTS OF UNIVARIATE logistic regression analysis are shown in Table 1. Of the 10 clinical factors, five factors, including testicular volume, serum concentrations of TT, FT, and inhibin B, and JS, were found to be significant factors influencing the success of microdissection TESE. The results of stepwise multivariate logistic regression analysis are shown in Table 2. This model showed age, FSH and JS to be the most influential factors. The regression coefficients of the optimum model were estimated, and the mathematical model predicting the probability of a successful outcome was as follows:

Table 1.

Univariate predictors of successful microdissection testicular sperm extraction outcome including Johnsen's score

Variable	Regression coefficient		P‐value
Age	b 0	−0.3177	0.8238
	b 1	−0.0014	0.9738
Testicular volume	b 0	−2.0200	0.0004
	b 1	0.1600	0.0015
LH	b 0	−0.1880	0.6301
	b 1	−0.0250	0.5991
FSH	b 0	0.1560	0.7056
	b 1	−0.0210	0.1566
TT	b 0	−2.0785	0.0008
	b 1	0.4794	0.0030
FT	b 0	−3.3000	0.0050
	b 1	0.2000	0.0166
E2	b 0	−0.5975	0.2247
	b 1	0.0086	0.6123
PRL	b 0	−0.3072	0.2942
	b 1	−0.0034	0.8508
Inhibin B	b 0	−1.1340	0.0004
	b 1	0.0190	0.0024
JS	b 0	−2.8600	<0.0001
	b 1	0.8700	<0.0001

log P/1 − P = b ₀ + b ₁ × explanatory variable. E2, estradiol; FSH, follicle stimulating hormone; FT, free testosterone; JS, Johnsens's score; LH, luteinizing hormone; PRL, prolactin; TT, total testosterone.

Table 2.

Multivariate predictors of successful microdissection testicular sperm extraction outcome including Johnsen's score

Variable	Regression coefficient	Odds ratio	95% CI for odds ratio	P‐value
Intercept	−1.134			0.6054
Age	−0.144	0.87	0.74–1.01	0.0675
FSH	0.059	1.06	1.01–1.11	0.0185
JS	1.31	3.7	2.07–6.65	<0.0001

CI, confidence interval; FSH, follicle stimulating hormone; JS, Johnsen's score.

P = (1 + exp[0.144 × patient age − 0.059 × serum FSH concentration − 1.310 × JS])⁻¹

The proportion of variation explained by this model was 0.87.

The ROC curve, drawn according to the predicted probability of a successful outcome by the optimum logistic model, is shown in the Figure 1. A predictive probability of 49.7% at the point where the ROC curve and the diagonal line from upper left to lower right intersected was selected as the best cut‐off value. Sensitivity, specificity and the AUC^ROC evaluated at the best cut‐off value were 78.0%, 76.3% and 0.83, respectively.

Figure 1.

Receiver operating characteristic analysis of the predicted probability of a successful outcome calculated by the equation obtained from multivariate logistic regression analysis. A predicted probability of more than 49.7% is the best cut‐off value. Sensitivity, specificity and the area under the curve are 78.0%, 76.3% and 0.83, respectively.

DISCUSSION

UNIVARIATE ANALYSIS SHOWED JS to be the most useful predictive factor for successful microdissection TESE (P < 0.0001, Table 1). Many studies have shown a relation between testicular histopathologic findings and testicular sperm retrieval by TESE. ¹² , ¹³ , ¹⁴ , ¹⁵ The accuracy of predicting TESE success was reported to be 87% if at least one spermatozoon was observed during histopathologic examination of a random testicular biopsy specimen. ¹² Histologic findings are generally the most useful predictive factor for successful TESE. It seems, however, too optimistic to expect to predict the outcome of TESE by testicular histology alone because many factors can affect and be affected by the complicated process of spermatogenesis. Therefore, multivariate analysis with preoperative factors, including testicular histology, is needed for prediction of a successful outcome. One interesting reported multivariate analysis revealed the importance of testicular histology; histologic findings were found to be the most predictive factors analyzed. ¹⁶ We found JS to be the most influential factor for prediction of successful microdissection TESE also by multivariate analysis (P < 0.0001, Table 2). Therefore, preoperative testicular histologic findings based on diagnostic biopsy or failed conventional TESE can help predict microdissection TESE outcome, although it is still controversial whether invasive examination, such as testicular biopsy, should be performed because it can cause inflammatory changes, hematoma, parenchymal fibrosis or permanent devascularizaion of the testes. ¹⁰

The result of our multivariate analysis showing that the serum FSH concentration also can help to predict outcome is reasonable, because the serum FSH concentration is generally correlated inversely with impaired spermatogenesis, and elevated FSH levels have been associated with a high probability of absence of spermatozoa. ¹⁷ , ¹⁸ Our previous study also showed that the serum FSH concentration was one of the factors necessary to predict TESE outcome. ⁹ However, the serum FSH concentration is related inversely to the total number of testicular germ cells present, not to sperm count or production. Thus, it is actually difficult to predict the most advanced stage of spermatogenesis, equivalent to TESE outcome, by the serum FSH concentration alone. ¹⁶ , ¹⁹ , ²⁰ , ²¹ Our univariate logistic regression analysis data also showed no significant predictive value in the serum FSH concentration alone. We suggest that the serum FSH concentration is better used as a predictive factor associated with JS, as shown in the present study.

It is surprising that age can also affect the outcome of microdissection TESE. Because patients included in the present study were all diagnosed with idiopathic NOA, spermatogenic impairment seemed to be congenital and unrelated to the aging process. However, age‐related changes in human testes have been generally reported to include reduced testicular size, impaired spermatogenesis, reduced tubular length, increased thickness of tubular boundary tissue, sclerosis, focal mononuclear orchitis, dilatation of the rete testis, and a high incidence of azoospermia. ²² Weight of the testicular parenchyma, proportion of the testis occupied by seminiferous epithelium and the volume of seminiferous epithelium are significantly reduced in aged men. ²² A study of the balance of apoptosis and proliferation of germ cells showed that accelerated apoptosis of primary spermatocytes might account in part for the mechanism of germ cell loss in aging men. ²³ Pronounced apoptosis events in germ cells could lead to oligospermia or azoospermia. ²⁴ , ²⁵ Thus, the aging process can worsen the seminiferous condition, even for patients with NOA. It was reported that a patient with Yq microdeletions in the Y chromosome, which include deletion of azoospermic factor (AZF) causing spermatogenetic impairment, was able, from age 25–38 years, to father four children but was later found to be spontaneously azoospermic. ²⁶ A similar case of severe oligospermia with AZF microdeletion, later manifesting as azoospermia, was recently reported. ²⁷ The patient underwent two testicular biopsies 1 year after diagnosis of azoospermia. The mean JS of the second biopsy (2.1) was lower than that of the first one (4.0). This finding might indicate progressive tubular depletion of germ cells, although the first biopsy might have worsened the testicular condition. Also, a single biopsy might be unrepresentative of the whole testicular picture because of the histologic heterogeneity of patients with NOA. Taken together, these findings and our result showing age as one of the most influential factors suggest that the aging process can affect spermatogenic impairment, which seems to be a congenital disorder, even for patients with NOA.

We previously reported a model calculating the predicted probability of a successful outcome for microdissection TESE, that was based on multivariate logistic regression analysis of non‐invasive factors without testicular histology: P = (1 + exp[5.201 – 0.048 × serum FSH concentration − 0.449 × TT concentration − 0.021 × serum inhibin B concentration])⁻¹. ⁹ The proportion of variation explained by that model was 0.77. Sensitivity and specificity evaluated at the best cut‐off value were 71.0% and 71.4%, respectively. The new model for predicting the probability of a successful outcome obtained in the present study is more useful because the proportion of variation explained by this model is very high at 0.87, and sensitivity and specificity are 78.0% and 76.3%, respectively, when a predictive probability of 49.7% is used as the best cut‐off value. If patients, who consider microdissection TESE, can represent their testicular histology by either preoperative testicular biopsy or prior failed conventional TESE, our new model is very useful for them. We conclude that the new model in the present study can help clinicians to recommend re‐TESE with microdissection method to patients who failed conventional TESE, although another methodology to have a better negative predictive value to avoid unnecessary surgical interventions is also useful. Finally, we suggest that younger age patients would be beneficial for microdissection TESE, as the aging process can affect spermatogenic impairment, even for patients with NOA.